Martin Egger

Hey,

I've developed a wedging setup for our RoyalRender farm that runs primarily on ROPs, but for the contact sheet creation I do rely on PDG. My current issue is that I'm now overloading my system with a large number of work items (24k image files for 24 wedges with 1000 frames each). Simple operations like string splits and so on just lead to Houdini freezing indefinitely, before I even get to the imagemagick part. RAM or CPU don't show a spike.




I was wondering if that is just in and out of itself a ridiculous number to execute, or I'm missing some settings. I thought PDG was supposed to be scalable like this. The setup does work on smaller batches with like 300 frames*16 wedges on other shots perfectly, but even there it does take quite quite some time to compute which I would hope it somehow wouldn't.

I thought about concatenating work items into batches where one work item stores each image path as an array, but the system is quite fragile and I'd probably need to rebuild it from scratch and lose a lot of the simplicity of tops.

Thanks for your help,
Martin

Super old thread, but I thought I'd place this here for completeness…
This site now lists all these attributes:
https://www.sidefx.com/docs/houdini/model/attributes.html [www.sidefx.com]

jsmack
Did you try creating a package? That's the best way to associate a plugin with a specific version of Houdini.

axiom.json
{
    "enable": false,
    "path": "C:/USERS/PUBLIC/DOCUME~1/Axiom/${HOUDINI_VERSION}"
}

Hey Jonathan,
off topic package syntax question here: Can you elaborate the naming you just put in there with the “~1”? Does that evaluate into the first matching folder with “DOCUME” like tabbing in the shell on a linux machine? I was curious as the Houdini textport evaluates paths similarly.
Just ever confused using windows and having unix syntax in some places for Houdini.

Midphase
How does it compare to EmberGen?

https://jangafx.com/software/embergen/ [jangafx.com]

On that account, I believe the new 18.5 minimal solver (using a newer, more efficient version of VDBs) is similar, but Axiom is more production capable with the sparsity and ability to switch with the same algorythms, as the new minimal solver does not support that (yet).
Also with the Houdini minimal solver, I've experienced the promised switch from GPU to CPU NOT to be without change, so it does not match exactly, which is a shame - but I might be wrong. Concerning that, does anyone have experience with Axiom switching for final highres export?